OpenAI just launched GPT-4o, a multimodal model capable of reasoning about text, voice and video in real time.
The demo videos are really impressive about the reasoning, but one has to be be really scared about the personality of the 'persona' in the machine.
It is really over-positively-enthusiastic about everything and about drying paint. It is the hyperactive overly eager friend that tries so hard to please that it becomes completely obnoxious. And it is chatty, trying to convey emotion to every answer, to every interaction. What a pain to interact with.
The researchers from Open AI where obviously enamoured with their creation in the videos, but the 'persona' they created is nothing more than a Frankenstein striving to much to become human.
Someone had to choose that style for the persona—I don't want to start thinking that this personality is a byproduct of the training data. The latter would be very unsettling.
- Beamer slides with markdown and pandoc
- HTMX resources
- Apple buys Pixelmator
- Working on a blog generator in bash
- Reading texts on paper beats reading texts on a computer screen
- Eureka Labs, AI for Education
- Reading on Darwin Machines
- You are not a Parrot, and other AI stories
- On Rags - Retrieval Augmented Generation
- Assange is free
- Editors, Editors
- Adding something between the lines
- Some Games I Like
- AI Aesthetics
- Random Number Generator, Anyone
- veo, i veo google
- gpt-4o, the crazy over helpful assistant
- teaching physical computing notes
- making checklists and ai
- refactoring old code scares me
- quem não lê é menos livre
- haiper video generation
- qualcomm ai hub
- New macs are rubbish
- I'm giving Zed a try
- where has my disk space gone?
- stealing this one, on how to study
- styles for clean bw slides in marp
- array based languages
- all 2023 posts
- added a rss feed
- code snippet for highlight.js in markdown pages
- is this the only css you'll ever need?
- things I'm (re)learning as i play advent of code 2023
- spell checker for shell scripts
- a simple ps1 bash prompt
- axtel and epstein's levels for agent based performance
- arduino tip120 demo code for classes
- blogs without server side rendering
- data science handbook
- grimm's odd standard protocol for describing individual-based and agent based models
- models of creativity
- ants are amazing - what about organizations - reading
- creating indices with tree
- m3, m3 pro and m3 max
- i have to many rss feeds in my reader
- finder, explorer, nautilus, rox, ... spacedrive
- marginalia in the modern digital world. is it possible?
- computer related stuff---how these machines work
- disable macbook air autoboot when dis/connecting to power or open lid
- red led at 13, blue led at 12 - police lights
- the wei (web environment integrity) api proposal from google is dangerous and should not go forward.
- my approach to managing scratch projects with bash scripts
- avoid long urls extending past the margins of text in latex with xurl
- how we learn and how to organise a reading inbox
- I am back to social networks, and it is mastodon.
- organising stuff is hard until it is not
- writing slides structure from the topics slide in marp
- a simple css trick 4 dithered images
- readings on strange programming, art and electronics
- gpt4 experiments - sparks of agi
- interesting openstreet use for studying hospital accessibility
- a simple js range one-liner
- o fim das trotinetes de aluguer?
- still reading about ai and gpt and what is next in this space
- the ai races for march 22:
- two main developments in the ai generators world
- and it goes dark
- fuzzy logic shell prompt alias
- svelte link dump.
- senhor clemente, que oportunidade perdida para o arrependimento.
- toggling light bulb problem
- no arrendamento, quem se lixa é quem cumpre e já aluga
- in the slow movement you find great pearls of wisdom.
- beja e alverca
- websites que funcionam apenas em modo texto
- small is beautiful
- tools for modern research
- and we are in 2023
- reading
- ai
- me